Some of this discussion about the merits of LISP-ALT and of other
schemes is only happening on the RRG list and so is not on the other
lists crossposted in this reply.

  http://www.irtf.org/pipermail/rrg/


Hi Brian,

You wrote, in part:

> 1. Large companies with their own international networks will
> surely never adopt a solution which doesn't allow them to have
> a straightforward, consistent and 100% internally managed
> addressing plan.

I agree.  In my view, this means they want and need portable address
space.  I think placing "portable" anywhere near "address" in a
sentence makes some peoples' teeth itch, including especially
yourself!  But I think this is what networks want and need: stable
public address space they retain for year after year, and can use via
any ISP, ideally being able to split it up very finely between
multiple ISPs and to change which ISP they use it with - quickly and
with little cost to themselves or other people.

The core-edge separation schemes LISP, APT and Ivip all provide
portable address space, like PI space, but not handled directly by
BGP, not obtained in necessarily big chunks and probably obtained
from some company or other organisation, rather than directly from an
RIR.

Also, this new kind of space (I call it Scalable PI space) can be
sliced up much more finely by the scheme's mapping system, down to
the individual IP address, compared to the administrative limits
placed on BGP's finesse, such as /24 for IPv4.  (Ivip only goes to
/64 resolution for IPv6.)  This space will have initial and ongoing
costs, but it won't involve BGP expertise, BGP routers, being an AS etc.


> 2. They also won't adopt a solution in which local customers
> experience world-crossing delays for accessing local sites.
> They will use DNS tricks, redirects, CDNs, and overlay routing
> to get round this.

I agree.  This rules out the use of a core-edge separation scheme
with a global query server system (LISP-CONS, LISP-ALT and TRRP).
There are ways of caching the mapping closer to ITRs, but this raises
problems.


> It's to be hoped that loc/id solutions such as LISP will be viewed
> as a help in this game rather than as a new enemy.

I tend to think that HIP is a real locator ID separation system.  I
don't think of LISP, APT or Ivip in these terms.

I think "Core edge separation scheme" is a better general term for
LISP, APT, Ivip and TRRP.  (Maybe RANGER too.)

But I agree - if LISP-ALT only provides a global query server system
and if it became "The Routing Scaling Solution" for solving the
routing scaling problem, then I think many people would consider the
delays and fragility of a global query server system sufficient
reason not to adopt the scheme.  Yet, in order to solve the routing
scaling problem, we need pretty much all end-user networks, large and
small (who want multihoming, portability etc.) to adopt the one scheme.

  - Robin

_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to